Aberrant voxel‐based degree centrality and functional connectivity in Parkinson's disease patients with fatigue

Abstract Aims The study aimed to investigate alterations in the inherent connectivity pattern of global functional networks in Parkinson's disease (PD) patients with fatigue. Methods Eighteen PD patients with fatigue (PD‐F), 20 PD patients without fatigue (PD‐NF), and 23 healthy controls (HCs) were recruited and analyzed by the voxel‐wise degree centrality (DC) and the seed‐based functional connectivity (FC) analysis. Meanwhile, the surface‐based morphometry (SBM) analysis was also commanded to explore the structural alternations among groups. Results PD‐F patients displayed reduced DC values in the left postcentral gyrus relative to PD‐NF and HCs groups, while increased DC values in the bilateral precuneus compared to HCs. Simultaneously, altered DC value in the left postcentral gyrus negatively corresponded to the mean fatigue severity scale (FSS) in PD‐F patients. Additionally, the receiver operating characteristic (ROC) curves uncovered that the reduced DC value of the left postcentral gyrus could discriminate PD‐F from PD‐NF and HCs groups. Our FC analysis further revealed that altered FC was located predominantly in the sensorimotor network in the PD‐F group. Moreover, we discovered no statistically significant differences between the three groups concerning cortical thickness. Conclusion Our findings indicated that the altered functional connectivity in the sensorimotor network centering on the left postcentral gyrus and the bilateral precuneus might be the potential pathogenesis of PD with fatigue.

to solve in PD. Although fatigue is considered to be caused by the complicated interaction between the latent disease processes, peripheral control systems, central control systems, and environmentrelated elements, 3,4 the exact pathophysiological mechanism of PD with fatigue remains unknown.
Concerning the pathophysiological mechanism of PD with fatigue, a growing literature has proposed that the destruction of the striato-thalamo-prefrontal loop may play a role in PD with fatigue. 3,[5][6][7] Moreover, a functional imaging study showed that among patients with PD who were drug-naïve, altered network connectivity in the sensorimotor net and default mode network (DMN) might be affiliated with fatigue. 8 Another piece of evidence uncovered that fatigue in PD was correlated with the terrible attenuation of sensory signals from the somatosensory region to advanced motor systems. 9 Thus, we inferred reasonably that PD patients with fatigue had atypical changes in the intrinsic connectivity pattern of brain-wide functional networks.
Resting-state functional magnetic resonance imaging (rs-fMRI), offering a different perspective on neural network changes in PD, has been widely used in recently exploring the mechanism of fatigue in PD. Degree centrality (DC), a voxel-level measurement of network connectivity, is one of the persuasive and reliable methods among nodal network metrics. 10 DC distinguishes the node with the most connection by calculating the number of direct links to others to determine the relative importance of nodes within the network, which has been certificated anomaly in PD and related complications. [10][11][12][13] Therefore, we adopted the voxel-wise DC approach to inspecting intrinsic disconnection patterns in the fatigued brains of PD patients.
Furthermore, we probed the relationships between anomalous DC values of the cerebrum and fatigue severity in PD fatigue patients.
We also conducted seed-based functional connectivity (FC) analysis and surface-based morphometry (SBM) analysis based on regions that exposed significant DC alterations to explore possible damaged networks and cortical thickness diversities of PD patients with fatigue.

| Subjects
Forty-two right-handed patients with idiopathic PD, drawn from the outpatient clinic of Neurology Department of the First Affiliated Hospital in Nanjing Medical University, were initially included in our study. Inclusion criteria were: (1) all patients obeyed the Movement Disorder Society clinical diagnostic criteria for PD 14 ; (2) all patients were between the age of 40 and 80 years old; (3) all patients followed the modified Hoehn and Yahr (H&Y) rating lower than the fourth stage; (4) all patients took medicine steadily for at least 1 month.
Below were the criteria for exclusion: (1) patients with unclear PD diagnosis; (2) patients with severe neurological and psychiatric diseases who could not cooperate with researchers, and those who struggled with serious respiratory, cardiovascular, and organic brain diseases; (3) patients with MRI contraindications; (4)

| Clinical assessment
Parkinson's disease patients were categorized with fatigue (n = 18, average FSS >4) and without fatigue group (n = 20, average FSS ≤4) by fatigue severity scale (FSS). FSS, a 9-item questionnaire, asks participants to mark the severity and impact of fatigue on daily life on a 1-7 scale (strongly disagree to strongly agree) and has been widely used in PD fatigue-related research. 15 In addition, the modified H&Y stage and Unified Parkinson's Disease Rating Scale III (UPDRS-III) were used to evaluate the stage of disease and severity, respectively.
MMSE was used to measure the overall cognitive state, while HAMD and HAMA to assess depression and anxiety, ESS to evaluate sleep disorders, 16 and AS to estimate the apathetic state. 17 Moreover, we computed the levodopa equivalent daily dose (LEDD) for patients involved to judge the use of dopaminergic drugs. 18 Notably, all clinical scales and MRI data acquisition were carried out at least 12 h after the drug's withdrawal to avoid possible pharmacological interferences in our experiment.

| MRI data acquisition
We performed MRI examination using the Siemens 3.0-Tesla signal scanner (Siemens Medical Solutions) equipped with an eightchannel phased array head coil. During the scan, foam padding and earplugs were prepared to restrict the motion of the head and curtail scanner noise. A further requirement was that all subjects kept still, shut their eyes, and did not think about anything special during the examination. T1-weighted and sagittal 3D magnetization-prepared rapid gradient echo (MP-RAGE) sequences obtained high-resolution brain structural images with parameters as follows: repetition time

| Preprocessing of fMRI data
Imaging data were preprocessed and analyzed on DPABI software (http://www.restf mri.net/forum/ dparsf). Steps were divided into the following aspects: First, to eliminate the interference of unstable magnetic field and surrounding environment, we discarded the image data of the first 10 time points and corrected the remaining 230 images for the slice timing and head motion (Friston 24 parameter). One Institute of Neurology (MNI). We resampled the functional images with a resolution of 3 × 3 × 3 mm 3 , and used the Gaussian kernel (full width at half-maximum, FWHM = 6 × 6 × 6 mm 3 ) for spatially smoothing. Finally, bandpass filtering (0.01-0.08 Hz) was carried out, and the linear trend was removed. In fact, we also removed several noise covariates, including white matter noise signal, cerebrospinal fluid signal, and six head motion parameters obtained through head motion correction.

| Voxel-based DC analysis
Degree centrality data analysis is a voxel-level assessment based on DPARSF. Pearson's correlation coefficient matrix was generated by calculating the temporal correlation between a gray matter mask voxel and all other voxels within the brain. Then, to eliminate the low time correlation interference caused by signal noise, we set the threshold of Pearson's correlation coefficient to r > 0. 25. 10,19 Our study utilized binary DC values within the brain network to conduct subsequent statistical analyses. Afterwards, the fisher-z transformation was served to convert the voxel-wise DC values into a z-score graph to enhance normality. Ultimately, all individual DC graphs were spatially smoothed together in the preprocessing stage (FWHM = 6 × 6 × 6 mm 3 ).

| Seed-based FC analysis
Seed-based FC analysis was conducted based on the brain regions with significant DC changes (compared PD-F with PD-NF group) and correlated with the mean FSS. We selected and defined the clusters (brain regions) as regions of interest (ROIs) for seeds. Before FC analysis, we first executed spatial smoothing (FWHM = 6 × 6 × 6 mm 3 ).
Next, for obtaining FC maps, we extracted ROIs time courses by averaging the time series of all voxels within the seed region and calculated the correlation between the ROIs and each voxel's average time series within the brain, including all voxels in the ROI itself.
At last, we used Fisher's z transformation to convert FC maps into z-score graphs to upgrade normality effectively.

| Surface-based Morphometry analysis
Surface-based morphometry (SBM), carried out with CAT12, an SPM12 extension with the default pipeline (http://dbm.neuro.unije na.de/cat), was used for the cortical thickness analysis. Here, the distance measurement based on projection is employed to calculate the central surface, cortical thickness, and other indicators in barely one step. 20 3D T1-weighted images were segmented to WM, GM, and CSF automatically; spherical registration to an MNI template space was applied; extracted and produced four SBM data, including cortical complexity, cortical thickness, and so on; resampled and smoothed surface as mentioned earlier by a 15 mm FWHM Gaussian kernel.

| Statistical analysis
Data analysis was carried out based on IBM SPSS statistics 25.0 (SPSS). The Shapiro-Wilk test was performed to determine the data's normality. For normally distributed variables, one-way analysis of variance (ANOVA) and two sample t-test were adopted. For data that exhibited non-normally distribution, we used non-parametric test, including the Chi-square test, Mann Whitney U-test, and Kruskal Wallis test to compare sociodemographic and clinical data.
Bonferroni correction was also performed. Statistical significance was defined as p < 0.05.
The statistical analysis module of DPABI software was operated to calculate and analyze the DC values of the three groups.
We used the one-way analysis of variance (ANOVA) to detect significant DC differences among the three groups, along with the covariates of sex, age, education, and mean FD. Based on the results of ANOVA, we subsequently performed a two-sample post hoc ttest with covariates described above between each pair among the three groups. All statistical significance complied with voxel-wise p < 0.001 and a cluster-level p value <0.05 after AlphaSim correction between groups.
Additionally, we extracted brain areas with varying DC values between PD subgroups as ROIs to perform FC analysis in the whole brain. ANOVA and post hoc t-test were exerted to estimate FC differences between groups with sex, age, and education as covariates, combined with a voxel-wise uncorrected p < 0.01 and an AlphaSimcorrected cluster-wise threshold of p < 0.01.
In order to clarify the association between brain regions with different DC and FSS scores, we commanded partial correlation analysis to calculate the correlation between the average DC value of the selected ROIs and the mean FSS score of PD-F patients, taking the course of the disease, age of onset, LEDD, MMSE, HAMA, and HAMD-24 scores as covariates to eliminate the interference of confounding factors. Two-tailed p < 0.05 was considered a significant statistical difference. Subsequently, we utilized ROC curves to judge whether the extracted brain indicators could be regarded as PD-F identification features. The cut-off values' sensitivity, specificity, and area under the curve (AUC) were all reported. The optimal cut-off value was selected by maximizing the Jordan index.
Moreover, we also examined the Pearson's correlation between the average FC value and FSS score by SPSS 25.0 software to determine the correlation between the regions showing functional connectivity differences of ROIs and the severity of fatigue.
For SBM analysis, neural structure data was conducted on SPM12 and the CAT12 extension for inter-group analysis. A single factorial analysis model was used, with sex, age, and education as covariates.
Then, t contrasts were performed for inter-group comparison based on ANOVA analysis. The significant level selected was p < 0.05.

| Demographic and clinical characteristics
Clinical characteristics of the PD-F group, PD-NF group, and HCs group are listed in Table 1. The three groups observed no significant difference in age, sex, education, MMSE scores, and mean FD. We

| DC data and correlation analysis
As exhibited in Figure 1 and Table 2 Figure 2B and Table S1).

| FC data and correlation analysis
Regions showing significant differences in DC between the two subgroups of PD and associated with mean FSS scores were selected as the ROIs to detect the difference in brain network for FC analysis.  Table 3).
It is worth mentioning that the correlation analysis between the left postcentral gyrus and the left SMA-related FC network alterations negatively correlated with the mean FSS score of PD-F patients   Table S1 for more details.  Figure S1).

| SBM analysis
No statistical difference among the three groups for the cortical thickness at a threshold of FWE-corrected p < 0.05 was detected in this study.

| DISCUSS ION
Our  23 On this basis, we reasonably hypothesize that sensory input attenuation leads to altered connections between the sensory cortex and the higher-order motor cortex, such as the SMA, PMC, eventually resulting in fatigue in PD patients. FC analysis of PD fatigue in drug-naïve patients also found that fatigue was relevant to diminished connectivity of SMA in sensorimotor networks. 8 Correspondingly, a global FC analysis across brain networks revealed that lower connectivity in the central sensorimotor system seems to be linked explicitly to fatigue in multiple sclerosis patients. 24 Moreover, transcranial direct current stimulation (tDCS) and other related stimulation on the bilateral somatosensory cortex could effectively combat fatigue in patients with multiple sclerosis. [25][26][27] In line with previous reports, our correlation analysis exposed that fatigue severity in PD was related to the destruction of functional connections in the sensorimotor network. fatigue. 9 We speculated that this might be due to a compensatory response in the early stage of the disease, whereas compensation gradually transforms into the decompensation stage as the pathology progress, 28 which needs to be further verified by longitudinal research studies in PD fatigue.
The precuneus, an essential node of the DMN, 29 has been proven to be involved in visual perception, episodic memory retrieval, selfprocessing, and consciousness. 30 Our findings identified increased DC values of the bilateral precuneus in PD-F patients in comparison with HCs, which was deduced to compensate for the decreased sensorimotor network connectivity in PD-F patients by enhancing their functional connectivity. In fact, the precuneus, which serves as a compensatory region by enhancing signal synchronization, has been observed in cognitive impairment and freezing of gait in PD. 12,31,32 A previous fMRI study also reported that fatigued multiple sclerosis patients demonstrated decreased precentral and postcentral gyrus activation and increased precuneus activation, confirming that fatigued multiple sclerosis patients were likely to maintain overrecruit of the precuneus to adapt to the increased task demands. 33  of fatigue in PD, suggesting fatigue correlated with brain lateralization. 38 Unfortunately, we did not find differences in the structure of fatigued PD patients. Moreover, there is no clear and consistent evidence of functional or pathologic localization of lateralization in PD patients with fatigue.
Our research had some limitations. First of all, the sample size of patients included in this study was relatively small, affecting its representativeness. Nevertheless, our results were still trustworthy because they survived multiple comparisons correction, and we eliminated the interference of anxiety, depression, sleep, and so forth. Second, we used the clinical scale to evaluate the degree of fatigue in this study, lacking an objective measurement standard despite the availability and reliability of FSS that has been proved. 39,40 Third, this study was a cross-sectional study. At present, it might not be clear whether the findings of this study were adaptive changes or maladjustment. The current evidence was insufficient to clarify the causal relationship between changes in brain regions and fatigue, which should need further longitudinal research. Therefore, we may consider expanding the sample size for a future longitudinal follow-up study.

| CON CLUS ION
To conclude, our results showed that PD-F patients represented abnormal network connection density in the postcentral gyrus and aberrant FC in the sensorimotor network centering on the left postcentral gyrus and the bilateral precuneus. The findings also uncovered that these alternations were predictive factors of fatigue severity. In short, anomalous functional connectivity in the sensorimotor network system and the bilateral precuneus are closely related to fatigue in PD patients. The left postcentral gyrus is expected to be the potential imaging marker of PD fatigue.

AUTH O R CO NTR I B UTI O N S
AS, HZ, MG, LW, XC, CG, and HS jointly completed the research project and the statistical analysis. AS wrote the first draft of the article.
MG checked the article. HZ checked and revised the first draft. KZ and YY designed this study and revised the manuscript for intellectual content. All authors approved the final manuscript.

ACK N OWLED G M ENTS
We are very grateful to all the participants in this study.

FU N D I N G I N FO R M ATI O N
This work was supported by the National Natural Science Foundation of China (grant numbers 81671258 and 81901297).

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare that there is no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The original contributions presented in the study are included in the article/supplementary material; further inquiries can be directed to the corresponding authors.